]> git.baikalelectronics.ru Git - kernel.git/commitdiff
mm/migrate: convert move_to_new_page() into move_to_new_folio()
authorMatthew Wilcox (Oracle) <willy@infradead.org>
Fri, 13 May 2022 03:23:05 +0000 (20:23 -0700)
committerAndrew Morton <akpm@linux-foundation.org>
Fri, 13 May 2022 14:20:17 +0000 (07:20 -0700)
Pass in the folios that we already have in each caller.  Saves a
lot of calls to compound_head().

Link: https://lkml.kernel.org/r/20220504182857.4013401-27-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
mm/migrate.c

index ab46ad93af9f7071fa38a78a12fbdc85ffe271eb..1415ea25f7aa706856efd3a50ebf852d24a4aa38 100644 (file)
@@ -843,21 +843,21 @@ static int fallback_migrate_page(struct address_space *mapping,
  *   < 0 - error code
  *  MIGRATEPAGE_SUCCESS - success
  */
-static int move_to_new_page(struct page *newpage, struct page *page,
+static int move_to_new_folio(struct folio *dst, struct folio *src,
                                enum migrate_mode mode)
 {
        struct address_space *mapping;
        int rc = -EAGAIN;
-       bool is_lru = !__PageMovable(page);
+       bool is_lru = !__PageMovable(&src->page);
 
-       VM_BUG_ON_PAGE(!PageLocked(page), page);
-       VM_BUG_ON_PAGE(!PageLocked(newpage), newpage);
+       VM_BUG_ON_FOLIO(!folio_test_locked(src), src);
+       VM_BUG_ON_FOLIO(!folio_test_locked(dst), dst);
 
-       mapping = page_mapping(page);
+       mapping = folio_mapping(src);
 
        if (likely(is_lru)) {
                if (!mapping)
-                       rc = migrate_page(mapping, newpage, page, mode);
+                       rc = migrate_page(mapping, &dst->page, &src->page, mode);
                else if (mapping->a_ops->migratepage)
                        /*
                         * Most pages have a mapping and most filesystems
@@ -866,54 +866,54 @@ static int move_to_new_page(struct page *newpage, struct page *page,
                         * migratepage callback. This is the most common path
                         * for page migration.
                         */
-                       rc = mapping->a_ops->migratepage(mapping, newpage,
-                                                       page, mode);
+                       rc = mapping->a_ops->migratepage(mapping, &dst->page,
+                                                       &src->page, mode);
                else
-                       rc = fallback_migrate_page(mapping, newpage,
-                                                       page, mode);
+                       rc = fallback_migrate_page(mapping, &dst->page,
+                                                       &src->page, mode);
        } else {
                /*
                 * In case of non-lru page, it could be released after
                 * isolation step. In that case, we shouldn't try migration.
                 */
-               VM_BUG_ON_PAGE(!PageIsolated(page), page);
-               if (!PageMovable(page)) {
+               VM_BUG_ON_FOLIO(!folio_test_isolated(src), src);
+               if (!folio_test_movable(src)) {
                        rc = MIGRATEPAGE_SUCCESS;
-                       ClearPageIsolated(page);
+                       folio_clear_isolated(src);
                        goto out;
                }
 
-               rc = mapping->a_ops->migratepage(mapping, newpage,
-                                               page, mode);
+               rc = mapping->a_ops->migratepage(mapping, &dst->page,
+                                               &src->page, mode);
                WARN_ON_ONCE(rc == MIGRATEPAGE_SUCCESS &&
-                       !PageIsolated(page));
+                               !folio_test_isolated(src));
        }
 
        /*
-        * When successful, old pagecache page->mapping must be cleared before
-        * page is freed; but stats require that PageAnon be left as PageAnon.
+        * When successful, old pagecache src->mapping must be cleared before
+        * src is freed; but stats require that PageAnon be left as PageAnon.
         */
        if (rc == MIGRATEPAGE_SUCCESS) {
-               if (__PageMovable(page)) {
-                       VM_BUG_ON_PAGE(!PageIsolated(page), page);
+               if (__PageMovable(&src->page)) {
+                       VM_BUG_ON_FOLIO(!folio_test_isolated(src), src);
 
                        /*
                         * We clear PG_movable under page_lock so any compactor
                         * cannot try to migrate this page.
                         */
-                       ClearPageIsolated(page);
+                       folio_clear_isolated(src);
                }
 
                /*
-                * Anonymous and movable page->mapping will be cleared by
+                * Anonymous and movable src->mapping will be cleared by
                 * free_pages_prepare so don't reset it here for keeping
                 * the type to work PageAnon, for example.
                 */
-               if (!PageMappingFlags(page))
-                       page->mapping = NULL;
+               if (!folio_mapping_flags(src))
+                       src->mapping = NULL;
 
-               if (likely(!is_zone_device_page(newpage)))
-                       flush_dcache_folio(page_folio(newpage));
+               if (likely(!folio_is_zone_device(dst)))
+                       flush_dcache_folio(dst);
        }
 out:
        return rc;
@@ -1001,7 +1001,7 @@ static int __unmap_and_move(struct page *page, struct page *newpage,
                goto out_unlock;
 
        if (unlikely(!is_lru)) {
-               rc = move_to_new_page(newpage, page, mode);
+               rc = move_to_new_folio(dst, folio, mode);
                goto out_unlock_both;
        }
 
@@ -1032,7 +1032,7 @@ static int __unmap_and_move(struct page *page, struct page *newpage,
        }
 
        if (!page_mapped(page))
-               rc = move_to_new_page(newpage, page, mode);
+               rc = move_to_new_folio(dst, folio, mode);
 
        /*
         * When successful, push newpage to LRU immediately: so that if it
@@ -1261,7 +1261,7 @@ static int unmap_and_move_huge_page(new_page_t get_new_page,
        }
 
        if (!page_mapped(hpage))
-               rc = move_to_new_page(new_hpage, hpage, mode);
+               rc = move_to_new_folio(dst, src, mode);
 
        if (page_was_mapped)
                remove_migration_ptes(src,